Some Say AI Is the Greatest Invention of All Time. I Don’t Get It.

The conversation about artificial intelligence remains absurd, hype-ridden, and utterly out of touch with actual material reality. I could have written that sentence in 2024, 2023, or 2022, and it would have also been true. But it felt particularly true earlier this week, when America woke to the news that the stock price of Nvidia, a Silicon Valley company responsible for a lot of our AI breakthroughs, had tanked because a Chinese start-up had succeeded in quickly and cheaply making comparable models.

Many, many powerful people have said that artificial intelligence is one of the most important human inventions of all time. My reaction to them is: Wow, these people must really enjoy shitting in the yard.

Here’s an important human invention: plumbing. Bringing fresh water from one place to another, and disposing of human waste via engineering. It wasn’t until the 1960s that a majority of American homes could do this, which means that the beginning of the Space Age overlapped with a period when most Americans couldn’t wash their hands whenever they wanted to. And as cool as launching satellites and orbiting the earth and traveling to the moon are, their practical impacts on human life pale in comparison to modern plumbing.

Anyone who cites AI as the pinnacle of human ingenuity, above plumbing, should try spending a month without the latter. You have to piss outside. If you want to wash your hands after doing so, your problems multiply. You have to walk to a well, if you can find one, to get (hopefully clean) water, and then you have to heat it up on your stove if you want it hot. You can’t shower, and taking a bath would be a remarkably laborious process. Before mass indoor plumbing, cholera, typhoid, gastrointestinal worms, scarlet fever, hepatitis, and more were massively harder to avoid. I don’t know you, but I feel considerable confidence in suggesting that your desire to avoid those diseases is greater than your attachment to ChatGPT.

You might call this the shitting-in-the-yard test—or the indoor plumbing test, for those who prefer to avoid vulgarity. The test requires you to compare the hype about a particular tech product to a brick-and-mortar change wrought in the last century. Is Zoom really a bigger part of your life than food refrigeration, a technology that has saved untold millions of lives over the decades by dramatically reducing deaths from foodborne illness? Is cloud storage really a bigger deal than infant vaccines, which save six lives a minute? Is AI more important than the bowl? By which I mean, bowls. To put food in. To eat out of. Try and spend the rest of your life without ever using another food container and get back to me about whether generative AI is more important. Food containers are inventions, too!

What can AI allow us to do today that we couldn’t do without AI? Watch this Apple Intelligence advertisement. It is about a dumb and lazy white-collar worker named Warren who spends his days in the office playing with a tape dispenser. He spends 30 seconds writing a very bad email, then uses AI on his iPhone to translate it into a more professional, formulaic 50-word message that sounds fine. When Warren’s boss receives it, he is flabbergasted. That is how low his expectations of Warren are. The explicit message of this ad is that the product being sold is for the dumbest people alive. And Apple is not the only company that’s selling AI by trumpeting its ability to shepherd the tragically stupid through life.

I struggle to see how fobbing off minor administrative tasks to software warrants the hype. I can tell you the social importance of replacing oil lamps with incandescent bulbs and of replacing incandescents with LEDs. I can’t tell you the social importance of allowing Warren to get away with being a shitty employee.

But you can understand the conundrum: AI is being sold with the most outsize hype of any development in my 43 years of life, but what AI can do right now is maybe automate a few dull tasks that afflict the white-collar worker. A little over two years since people declared the world forever changed by the release of ChatGPT, what large language models (LLM) can actually do is remarkably limited. They can access and synthesize information, but not better than an educated adult. For important tasks, almost anyone will choose to do that work themselves, especially given ongoing issues with the outputs of LLMs. Doing it yourself is how you get and stay smart. And when LLMs get as good at humans at writing formulaic emails. . . 

. . . so what? Before the invention of the airplane, we couldn’t fly. After the invention of the airplane, we could. In any plausible nearish future, what affordances can AI provide us that could not be produced via other means?

You can use AI to produce images from text, but the output is nothing that humans couldn’t already produce, and working with a graphic designer will be far more effective when it comes to actually implementing your vision. And anyway, for the vast majority of people, the ability to generate images from text falls firmly into the “that’s cool” category of minor amusements, not into the realm of real, practical consequence. Also, I note with some amusement that you can now pay to take classes in how to use text-to-image engines most effectively. This technology is supposed to be accessible, yet like so much of what’s valuable in our society, the actual skill to use it is paywalled. Moreover, if you trawl forums dedicated to these engines, you’ll find immense frustration even among their most devoted users; for every amazing image someone shares that came from an AI, there are dozens that came out borked and were discarded.

The ways in which AI could be really useful remain firmly in the realm of science fiction, like the idea it could eliminate money or end death. Yet the absolutely constant hype inflation never stops. Here’s someone named Ross Lazer claiming that the ability of an AI “agent” to order a pizza for you is as transformative as the automobile. To be clear: What’s not being referred to here is the ability to order a pizza online, which is an affordance so old that pizza is believed to be the first thing ever sold via the internet. No, what’s as transformative as the automobile—which utterly changed commerce, socializing, and our lived environments—is simply the ability to get a bot to do that simple, decades-old task for you. I find it profoundly easy to order a pizza online. Can it really be a socially optimal use of resources—immense amounts of money, manpower, and electricity—to create incredibly complex systems that can, with tons of training and eye-watering power costs, take that simple task off my hands?

I don’t get it.

Or maybe I do. The ultimate source of the hype is obvious: money. Lazer is himself in the AI industry, so he has a vested interest in believing—or pretending to believe—that telling an “agent” on the computer to order a pizza is categorically more impressive than, uh, using a computer to order a pizza. The hype pays his bills. On this podcast episode, Kevin Roose of The New York Times keeps defending the outlandish claims of AI companies by insisting that people in the AI industry feel the same way. “They’re very sincere,” he says. To which I would say, You mean everyone whose stock holdings, and thus net worth, are directly related to AI hype is in agreement about AI hype? You don’t say! Sam Altman says there is no slowdown in improvements to LLM-based AI systems; his wealth is directly tied to public perception of whether there is a slowdown or not. These are not unrelated phenomena. What I don’t understand is why the media, so often, justify AI hype by credulously reporting what people in the industry are saying about it. Frito-Lay says that Doritos are snacktacular too! But I don’t take their word for it.

I struggle to see how fobbing off minor tasks to software warrants the hype.

Silicon Valley really needs AI to work, financially. It’s been looking for a new market to monetize for some time. Social media companies have relentlessly tried to wring every last cent out of their networks, and even so there are indicators that social media is a declining economic phenomenon. (Does “the next Facebook” sound like a remotely sexy investment to you?) Smartphones have become a technology that people don’t feel they have to upgrade every year anymore, while virtual reality has not lived up to its hype, at least financially. For years now, the tech industry has believed in limitless growth, but with so many mature product categories and saturated fields, right now one of the only areas of potentially explosive growth in Silicon Valley is AI—whether it’s actually useful or not. And the fact that Nvidia just lost almost $600 billion in hypothetical worth because a company in China produced an LLM system as powerful as the best offered by Silicon Valley, for a lot less money, is a sign that here, “worth” is an idea rather than a reality.

There will be some gradual improvements to productivity and time management from these systems, just as with routine software development. But whatever the hype machine says, I don’t think artificial intelligence will impact average quality of life anywhere near as much as changes in distribution of wealth, and in matters of war and peace. And I’d bet money that the technological growth most likely to actually revolutionize some aspect of human existence is not AI but biomedical science, probably genetic engineering specifically. In the field of IVF, for example, very likely developments in near-future technology would have immense consequences if realized.

I think there’s something else besides economics going on with the AI hype, though: People badly wish to escape mundane reality and all of its grinding indignities, and artificial intelligence—or those who market it, at least—has tapped into that desire. To those people, I say: in 10, and 20, and 30 years, you’ll still have to do chores that you hate, and if they find some way to automate away a particularly onerous chore, there will be some other petty task to take its place. That’s how human life works. You’ll still have to stand in impotent resentment while you wait for a subway train that will arrive already stuffed with too many riders. And if they invent the teleporter, then you’ll find other reasons to feel bored and annoyed. The good news is that for those who aren’t impoverished or seriously disabled, and who live in the developed world, there’s a great deal of opportunity for frequently sunny and generally comfortable lives that feature loving relationships. You will always be unsatisfied, but it will still be enough.